How to use the Stage Design Method agent skill

What the stage-design skill makes the agent do

The skill sets the order in which an agent builds a new classroom. The agent proposes a page plan, with a title, type and brief for each page, and confirms it with you before creating anything. It then creates the stage, sets the roster of teachers and agents, generates each page in order, adds narration audio, and checks the finished deck.

The skill also covers two recovery situations: retrying failed page generations, and re-synthesizing narration audio after the narration text has been edited. A stage stays unfinished until every settled page is persisted and every speech action has audio.

• Plans the classroom page by page: slide, quiz, interactive or pbl.
• Asks you to confirm the plan, unless you already gave a complete page structure or told the agent to go ahead.
• Binds each teacher and agent to a voice from the list of available voices.
• Generates pages one at a time, in ascending order, adding audio after each page.
• Verifies the stage with list_scenes and read_stage before reporting it done.

Install the stage-design skill

Download the ZIP from this page and unzip it. The skill is a single file named SKILL.md. Place the unzipped folder in one of two locations: ~/.claude/skills/ for all your projects, or a project's .claude/skills/ folder for one project only.

If you use the Claude apps instead, upload the skill under Settings → Capabilities → Skills.

• Personal install: unzip into ~/.claude/skills/.
• Project install: unzip into the project's .claude/skills/ folder.
• Claude apps: upload under Settings → Capabilities → Skills.

What the skill needs to run

The skill only works inside an OpenMAIC runtime that provides its stage tools: create_stage, set_roster, generate_scene, generate_tts, list_scenes, read_stage, patch_stage, ask_user, list_voices and register_voice. Without those tools, the instructions have nothing to call.

Narration audio also needs a voice provider configured in the runtime. The skill refers to voices as providerId::voiceId pairs, and the agent binds them from the list that list_voices returns.

• OpenMAIC runtime with the stage tools listed above.
• Voice provider configuration for narration audio.
• Scripts: none. Network: the skill itself does not use the network.

How to trigger the skill with natural requests

Ask for a new classroom or a rebuilt one in plain language, in the OpenMAIC environment. The skill is meant to load before generating or rebuilding any stage from scratch, whatever its subject.

If the stage already has pages, do not expect this skill to rebuild it. The skill states that such stages are edited under a different skill, pro-editing.

• "Build a new stage on photosynthesis for Year 9, with a quiz at the end."
• "Rebuild this classroom from scratch with the same topic but a new roster."
• "Make a four-page lesson on fractions and confirm the plan with me first."

Examples of a build from start to finish

A typical run starts with the agent proposing page titles, types and briefs. Once you confirm, the agent creates the stage, sets the roster, and generates each page in order. After each page succeeds, it generates that page's narration audio before moving on.

At the end, the agent checks that every settled page is saved, that every speech line has audio, and that no narration is blank. It then reports what was built and any gaps.

• Plan: each page gets a title, a type and a brief.
• Build: create_stage, then set_roster, then generate_scene and generate_tts for each page.
• Check: list_scenes, then read_stage on each page's actions.

Safety notes and tips

The skill runs no scripts and makes no network calls of its own. Its effects happen through the OpenMAIC tools it calls, so review the stage it produces before you publish it.

Keep the roster settled before generation starts. The skill notes that pages are written for the roster that exists when they are generated, so a late roster change leaves early pages out of step with it.

When you edit narration text, ask the agent to re-run audio generation on that page. Changed text with no new audio ships as a silent page.

• Settle the roster and page plan first.
• Re-synthesize audio after any narration edit.
• Ask for a gap report if a page fails to generate.